Obesity in childhood and adolescence appears to be an important predictor of adult obesity,1–3 and adult obesity is associated with high mortality. However, more long-term follow-up data on the relation between body mass index (BMI) in adolescence and in adulthood, as well as between adolescent BMI and adult mortality, is needed.4,5
Studies on persistence of obesity from adolescence into adulthood have suffered from several weaknesses; the studies have generally been small,1,6 with adult measurement in early adulthood only,7 or have had to rely on self-reported height and weight.7 Nevertheless, these studies have shown that obesity, to a large extent, persists from childhood and adolescence into adulthood. A more recent study, which followed approximately 11,000 individuals from birth to early adulthood, concluded that this persistence is largely explained by maternal weight or BMI.8
Because mortality in young adulthood is low, a large number of subjects and a long follow-up period are necessary to study the relation between obesity in adolescence and mortality. Only a few studies have investigated this relation.3,6,9–11 All-cause mortality has been shown to be elevated in overweight adolescents.3,10,11 One small study found an association of adolescent obesity and adult mortality after adjusting for adult BMI.6 One explanation that has been proposed for the effect of adolescent obesity on later mortality is the locus of fat deposition during adolescence.4
A workshop on childhood obesity convened by the International Obesity Task Force in 1997 concluded that BMI offers a reasonable measure of fatness in children and adolescents.12 Height and weight measurements are simple and inexpensive, and these have often been a routine part of health examinations. Hence, BMI has been calculated in many epidemiologic studies both among adolescents and adults.
Because adolescents are at various stages of maturation, age- and sex-specific growth curves are necessary to define obesity. Cole et al.13 proposed a set of such age- and sex-specific cutoff points, which were linked to adult categories of BMI (weight in kilograms)/(height in meters)2 for overweight (BMI 25–29) and obesity (BMI ≥30). In the United States, the Centers for Disease Control and Prevention (CDC) at the National Center for Health Statistics (NCHS) have created growth charts for children and adolescents up to the age of 20 years based on data from U.S. health examinations.14 The CDC/NCHS guidelines for adolescents suggest using age- and sex-specific BMI to identify adolescents at the upper end of the distribution as being “at risk for overweight” (BMI between the 85th and 95th percentile) and overweight (BMI ≥95th percentile).
The relation between BMI in adolescence and mortality has recently been explored in a large Norwegian cohort.10 The study included the complete cohort of adolescents measured during a tuberculosis screening program. The risk of death increased with increasing adolescent BMI. Some of the persons included in that study were also measured as adults in later population-based health surveys. The present study explored the relation between BMI in adolescence and in adulthood in a subcohort measured both at age 14–19 years and in adulthood. These persons were followed up with regard to death. Hence, it was possible to study independent effects of adolescent and adult BMI on the risk of death.
Height and weight were measured during 1963–1975 as part of a tuberculosis screening program in the general Norwegian population.15–17 This mass examination was compulsory for persons age 15 years and older, but height and weight were also measured in some persons less than age 15 years. The attendance was approximately 85% in persons above the age of 15 years.15 Previous reports have described the impact of adult height and weight on morbidity and mortality in this cohort.15,16,18 The relation between BMI in adolescence and mortality has also been explored using the complete cohort of adolescents (n = 227,003) measured during the screening program.10 Since 1973, height and weight have been measured in several other population-based health surveys in different parts of Norway.17,19 The attendance in the mid-1970s was 85–90%, but decreased to approximately 75% in the mid-1990s.17 Because of a unique 11-digit identification number assigned to all residents of Norway, it was possible to link separate measurements for each person.
Body weight (in kilograms) was measured using scales that were calibrated regularly and recorded to the nearest half-kilogram. Height was measured to the nearest centimeter. Measurements were excluded if they were not performed according to protocol (for example, the persons were wearing shoes) or if the woman stated that she was pregnant. We included all persons who were measured at least once in a screening for tuberculosis at age 14–19 years and again at least 10 years later. The lower limit of 10 years was chosen to ascertain that the persons were adults at the second measurement and to allow sufficient elapsed time between the measurements.
We excluded 281 persons registered with adolescent height more than 3 cm above adult height. This left 128,121 persons eligible for analysis.
In the population-based health surveys, the participants were asked to complete self-administered questionnaires. The questionnaires differed across surveys, and so although questions on smoking habits were included in all, the questions differed slightly. Thus, we classified the subjects only as “never-smoker” or “present or former smoker.”
It was possible to follow all persons except one from date of measurement until emigration, death or until June 30, 2002, using the unique personal identification number and the Death Registry at Statistics Norway.
We extracted the earliest BMI measurement at age 14–19 and the earliest measurement taken at least 10 years later for each person. We used multiple logistic regression models20 to explore the pattern of the relation between BMI in adolescence and adult obesity (BMI ≥30). BMI in adolescence was categorized following the guidelines from CDC/NCHS by using growth percentile curves from a U.S. reference population:14,21 low (<25th), medium (25th–74th), high (75th–84th), and very high (≥85th). From the web site of CDC/NCHS,21 we extracted percentiles for every 6 months from the age of 14–20 years. Complete percentile curves were then constructed by linear interpolation between these points. Growth curves from CDC/NCHS were chosen rather than constructing percentile curves from the study population to permit comparisons with other studies. Age at measurement in both adolescence (14–15, 16–17, and 18–19 years) and adulthood (24–29, 30–34, 35–39, and 40–54 years) were included in the model.
Multiple Cox proportional hazards regression models, with time since adult measurement as the time variable, were fitted to estimate relative risk of dying in different groups.22 BMI in adolescence was categorized as previously mentioned. The categorization of BMI in adulthood was <18.50, 18.50–22.49, 22.50–24.99, 25.00–27.49, 27.50–29.99, and ≥30.00. This categorization is more detailed than the recommendations from a WHO Consultation on Obesity in 1997,23 but is consistent with these recommendations.
The proportionality assumption in the Cox model was assessed by inspecting log-minus-log plots and the results from stratified analyses.
In addition to the variables included in the main models, we had information on area of residence at the time of each measurement (southeast, southwest, mid, and northern Norway), year of measurement, and smoking habits at the adulthood measurement (never-smoker and present or former smoker). The impact of these variables was explored both by inclusions in the models and by stratified analyses.
Analyses were performed separately for each sex. The results were presented as odds ratios (OR) of obesity in adulthood and relative risks (RR) of dying with 95% confidence intervals (CI).
The mean age of the 128,121 study subjects (61,522 men and 66,599 women; Table 1) was 17.0 years at the adolescent measurement (Table 2). Among adolescents, 64% of the boys and 65% of the girls had medium BMI, whereas 5% of the boys and 7% of the girls had very high BMI (Table 2).
Among men, the most common BMI categories at the adult measurement were 22.50–24.99 (32%) and 25.00–27.49 (30%) (Table 1). Among women, the most common adult categories were 18.50–22.49 (36%) and 22.50–25.00 (29%). Approximately 9% of men and 8% of women had a BMI of at least 30. The mean BMI was 25.6 among men and 24.2 among women. The mean time between adolescent and adult measurement was 23 years among both men and women (range, 10–34 years).
Persistence of Obesity
Table 3 shows the BMI at adolescent and adult measurement. Half the men with very high BMI in adolescence had BMI above 30 as adults, whereas only 5% of men with low or medium BMI had BMI above 30 as adults. Similar figures for women were 37% and 4%, respectively. Among obese (BMI ≥30) adults, 28% of men and 35% of women had very high BMI in adolescence.
The degree of persistence of obesity from adolescence into adulthood is illustrated in Table 4. The Hosmer-Lemeshow goodness-of-fit statistic was 12.2 in men and 10.1 in women on 6 degrees of freedom, giving P values of 0.06 and 0.12, respectively. The OR of obesity in adulthood increased steadily with increasing BMI in adolescence. Adjusting for other variables changed the estimates only slightly. The relative association between BMI in adolescence and adult obesity was roughly the same among men and women.
The persons in this study were followed for an average of 9.7 years (range 0–29 years) after the measurement in adulthood, comprising 1,239,523 person-years (Table 1). Among these, a total of 1682 deaths were observed. Mean age at death was 46.2 years in men and 46.5 years in women. By June 30, 2002, 126,013 persons (98.4%) were still alive and living in Norway, and 426 persons had emigrated.
Adolescents with high or very high BMI had a higher mortality than those with medium BMI (Table 5) after adjusting only for age at start of follow up (age at adult measurement). After adjusting for adult BMI, there was no excess mortality in men (relative risk [RR] = 1.1, 0.8–1.5), whereas the excess in women persisted (RR = 1.3, 1.0–1.7). Further adjustments for other variables did not change the results. Low adult BMI seems to be associated with an increased mortality in men but not in women.
Because BMI in adolescence and in adulthood are strongly correlated, we performed stratified analyses in which persons with high adult BMI were excluded from analysis of the association of adolescent BMI with mortality, or persons with high adolescent BMI were excluded from analysis of adult BMI and mortality. Hence, we looked at the independent impact of adolescent or adult BMI within a subgroup of the other. When including only persons with low or medium BMI in adolescence, the relation between adult BMI and mortality (Table 6A) was similar to that observed in the whole cohort (Table 5): among men, both low and high adult BMI was associated with an increased risk of death, whereas among women, increased risk of death was seen only for high adult BMI. When including only persons with adult BMI below 27.5, there was little effect of adolescent BMI in either men or women (Table 6B).
This study included persons whose height and weight were measured both in adolescence and in adulthood. The odds of being obese in adulthood increased consistently and rapidly with increasing BMI in adolescence. Adults who had been overweight adolescents had increased mortality compared with those of medium adolescent BMI (ie, BMI between the 25th and the 75th percentile in the U.S. reference population). After adjusting for adult BMI, there was no excess mortality in men, although the excess mortality persisted in women. On the other hand, the association between adult BMI and mortality was not influenced by adjustment for adolescence BMI.
One of the major strengths of this study was the large sample size, recruited from the general population. Adolescent weight and height measurements were obtained from the compulsory national tuberculosis screening program. Both in this program and in the later health surveys, measurements were performed in a standardized way. The follow up of the study subjects was almost complete until the end of follow up; of the 128,121 persons eligible for the study, 99.7% were registered as either being alive by the end of follow up or having died during follow up; 0.3% had emigrated.
Because of the large size of the dataset, it was possible to analyze many categories of BMI, height, age and year of birth instead of modeling these variables as continuous variables. The effects of various levels of the variables were therefore not “forced” into a specific pattern. The effect of adjusting for continuous confounders by the use of categorical variables has been questioned.24 Hence, we performed additional analyses, including age and year of birth, first as continuous variables and then as linear spline functions (data not shown). Similar results were obtained.
The predictability of overweight at age 35 years from adolescent BMI has been shown to increase with age during adolescence, from good predictability at age 13 years to excellent at age 18 years.1 We observed a strong association between adolescent overweight and adult obesity. The finding that 50% of men and 37% of women with very high BMI in adolescence had BMI above 30 as adults is roughly consistent with findings from other studies.3 Several risk factors for adult obesity have been reported previously; these include parental fatness, social factors, birth weight, timing of maturation, physical activity, and dietary factors.2,5,8
Because mortality in adolescence and young and middle-aged adulthood is low, a large number of persons and a long follow up are necessary to observe a sufficiently large number of deaths for precise estimates. In the present study, persons were followed for up to 29 years after the adolescent measurement, although they were followed for only 10 years on average after the adult measurement. Hence, even in this large cohort, the mortality data are limited.
Studies including BMI in adolescence have used different definitions of overweight and obesity.3 In the present study, we chose to use age- and sex-specific growth charts from CDC/NCHS to group adolescents by BMI. Another alternative was to use the growth charts from Cole et al.13 However, Cole et al. provided growth charts corresponding only to BMI of 25 and 30 at age 18 years. The CDC/NCHS growth charts made it possible to categorize in greater detail. Even so, the choice of growth charts was not important for the results. Similar results were obtained when we used the growth charts by Cole et al. (data not shown).
In a study from 1992, Must et al.6 followed 508 persons who were either lean or overweight during adolescence for 55 years. Leanness in adolescence was defined as BMI between the 25th and 50th percentiles, and overweight as BMI above the 75th percentile. An excess mortality was observed in men (but not women) who were overweight in adolescence compared with those who were lean. Adjusting for adult BMI decreased the relative risk only slightly. However, the adult BMI was calculated from self-reported height and weight. Errors in self-reported height and weight vary systematically with BMI.25 Hence, systematic misclassification, combined with a limited study size (n = 150 men), could have influenced the results. In the present study, excess mortality was observed in both men and women who were overweight in adolescence. Adjustment for adult BMI reduced the excess mortality in men, less so for women. Among persons with BMI below 27.5 in adulthood, there was no clear evidence of an effect of high adolescent BMI. Although the present dataset was large, it was not large enough to explore this issue more comprehensively, eg, with further stratification.
In the present study, we had information on smoking habits at the adult measurement. Like in the study of Must et al.,6 inclusion of smoking status to the regression model did not change the estimated relative risks.
In summary, this study showed that obesity in adolescence tends to persist into adulthood. Adolescent obesity is also connected to an excess mortality in adulthood, although this excess seems to be explained by obesity in adulthood, at least in men. High BMI in adolescence is predictive of both subsequent obesity and subsequent mortality.
We are grateful to those who, during almost 40 years, have collected the data used in the present study. These are persons connected to the former National Health Screening Service, The Nord-Trøndelag health survey (HUNT), The Hordaland health survey (HUSK), and The Tromsø study.
1.Guo SS, Chumlea WC. Tracking of body mass index in children in relation to overweight in adulthood. Am J Clin Nutr. 1999;70:145S–148S.
2.Whitaker RC, Wright JA, Pepe MS, et al. Predicting obesity in young adulthood from childhood and parental obesity. N Engl J Med. 1997;337:869–873.
3.Must A, Strauss RS. Risks and consequences of childhood and adolescent obesity. Int J Obes Relat Metab Disord. 1999;23(suppl 2):S2–11.
4.Dietz WH. Childhood weight affects adult morbidity and mortality. J Nutr. 1998;128:411S–414S.
5.Parsons TJ, Power C, Logan S, et al. Childhood predictors of adult obesity: a systematic review. Int J Obes Relat Metab Disord. 1999;23(suppl 8):1–107.
6.Must A, Jacques PF, Dallal GE, et al. Long-term morbidity and mortality of overweight adolescents. A follow-up of the Harvard Growth Study of 1922 to 1935. N Engl J Med. 1992;327:1350–1355.
7.Gortmaker SL, Must A, Perrin JM, et al. Social and economic consequences of overweight in adolescence and young adulthood. N Engl J Med. 1993;329:1008–1012.
8.Parsons TJ, Power C, Manor O. Fetal and early life growth and body mass index from birth to early adulthood in 1958 British cohort: longitudinal study. BMJ. 2001;323:1331–1335.
9.Dietz WH, Robinson TN. Use of the body mass index (BMI) as a measure of overweight in children and adolescents. J Pediatr. 1998;132:191–193.
10.Engeland A, Bjørge T, Søgaard AJ, et al. Body mass index in adolescence in relation to total mortality: 32-year follow-up of 227,000 Norwegian boys and girls. Am J Epidemiol. 2003;157:517–523.
11.Hoffmans MDAF, Kromhout D, de Lezenne Coulander C. The impact of body mass index of 78,612 18-year old Dutch men on 32-year mortality from all causes. J Clin Epidemiol. 1988;41:749–756.
12.Bellizzi MC, Dietz WH. Workshop on childhood obesity: summary of the discussion. Am J Clin Nutr. 1999;70:173S–175S.
13.Cole TJ, Bellizzi MC, Flegal KM, et al. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000;320:1240–1243.
14.Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al. CDC growth charts: United States. Adv Data. 2000;1–27.
15.Waaler HT. Height, weight and mortality. The Norwegian experiment. Acta Med Scand. 1984;Suppl 679:1–56.
16.Tverdal A. Body mass index and incidence of tuberculosis. Eur J Respir Dis. 1986;69:355–362.
17.Bjartveit K. The National Health Screening Service: from fight against tuberculosis to many-sided epidemiological activities [in Norwegian]. Nor Epidemiol. 1997;7:157–174.
18.Engeland A, Bjørge T, Selmer RM, Tverdal A. Height and body mass index in relation to total mortality. Epidemiology. 2003;14:293–299.
19.Bjartveit K, Foss OP, Gjervig T, et al. The cardiovascular disease study in Norwegian counties. Background and organization. Acta Med Scand Suppl. 1979;634:1–70.
20.Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley and Sons; 1989;1–307.
22.Cox DR, Oakes D. Analysis of Survival Data. London: Chapman and Hall Ltd; 1984;1–201.
23.World Health Organization Consultation on Obesity. Preventing and Managing the Global Epidemic: Report of a WHO Consultation on Obesity, Geneva, June 3–5, 1997. Geneva: World Health Organization; 1998;1–276.
24.Brenner H, Blettner M. Controlling for continuous confounders in epidemiologic research. Epidemiology. 1997;8:429–434.
25.Plankey MW, Stevens J, Flegal KM, et al. Prediction equations do not eliminate systematic error in self-reported body mass index. Obes Res. 1997;5:308–314.
© 2004 Lippincott Williams & Wilkins, Inc.